In the following I have converted lake level to metres below full supply level (so it’s a negative number, and positive numbers denote times when the dam is spilling), and this is plotted on the left y-axis. Currently I only have lake levels on a monthly basis for dates before 1 January 2000. After this date, I have used mean daily water levels computed from the raw 15-minute interval source data.
All water quality variables have had readings below the quantitation level (aka ‘detection level’) omitted, just so that we don’t get too distracted by them and have less cluttered plots. For the most part, these variables are denoted by circular symbols and represent mean values if > 1 reading was taken on the same day; the occasional near-vertical line is caused by samples taken a day apart - usually from different sites.
The time series plot are interactive in that there is a slider at the bottom of each which allows you to zoom in or out of parts of the time series. All plots start zoomed in on the period prior to the very high turbidity levelsof the mid-1990s. The interactivity allows you to point to particular items or lines and read of the values displayed in the legend at top right.
The horizontal red lines denote the upper limit of the ‘Medium’, ‘High’ and ‘Extreme’ risk bands respectively for water level in Woods Lake.
The following plot opens in the sequence before turbidities went throught the roof in the mid-1990s. Simply use the slider at the bottom to see what happened as the water levels dropped.
Note that the 1997 drought did not result in turbidities quite as high, and there seemed to be a longer time lag (perhaps?) in recovering to pre-draw down levels.
Nitrates and nitrites first, followed by the ‘total nitrogens’.
Similar thing for P:
The story for algal biomass is fairly similar, although I have been ruthless with the data in omitting all remotely suspect records and using the highest quantitation limit. There’s a bit more gardening required here because I suspect there have been a couple of different procedures used, and there’s a welter of negative values of for chlorophyll-a.
The motivation for working with Woods Lake was the perception that “water level drives turbidity, nutrients and, potentially, algal blooms”. So the flowing plots represent the same data from the time series, with time (and any autocorrelations), ahem, ignored. It might make more sense to lag the water quality variables with respect to the time of the recorded lake level, but what lag to use?
(This may become clearer when we examine the high-frequency platform data in a later document.)
In all the plots that follow, the year of collection encoded as a colour: darker blue = earlier in time.
Unsurprsingly, turbidity is greatest at lowest water levels, although the levels achieved in 2007 were much less than those in the mid-1990s.
Fairly similar patterns. For nitrate and nitrite, I have omitted a total of 3 points with concentrations >> 0.075 mg/L. We need to look at those later to see if they are valid values.
The first plot is for TP. Clearly the 1990s concentrations are higher at low water levels than the concentrations achieved later in the historical data, and this is accentuated when we log-transform TP in the second plot. Unfortunately, SRP is always at very low concentrations, so there is not much of a pattern in that graph.
And (yawn!)…
Again no lags in this stuff yet, but you’ve probably already guessed what these look like. Haven’t bothered coding the colours to year yet. Now that we start looking at these sorts of relationships, we can mess with analyses that include the BDL data (correlation coefficients for censored data and the like), but there’s a few more wrinkles to sort out first as well as some thinking about appropriate lags to use.